Just donated...

23 views
Skip to first unread message

mike503

unread,
Apr 8, 2008, 7:52:00 PM4/8/08
to highload-php-en
Everything left in my Paypal account ($19.12 or something)

I will send [much] more soon.

I just have to say the php_defines option saved me some hassle, I was
able to set heavy error reporting on a single FastCGI pool and not
across everything. So nice. So so nice!

I've already migrated everything I do to use php-fpm. It's awesome. I
cannot praise you enough, Andrei.

Quick question: when I do php-fpm reload or php-fpm restart, does it
physically kill the existing engines and someone could get a 502 bad
gateway during that little bit of time, or does it somehow do it in a
graceful fashion? It seems to restart so quick (and reload even
quicker) that it doesn't even trip up my healthchecks so far... I'm
wondering if it's doing it gracefully on purpose or it just works
really fast.

Andrei Nigmatulin

unread,
Apr 8, 2008, 8:24:27 PM4/8/08
to highloa...@googlegroups.com
On Wednesday 09 April 2008 03:52, mike503 wrote:
> Everything left in my Paypal account ($19.12 or something)
>
> I will send [much] more soon.

Thanks a lot !

> I just have to say the php_defines option saved me some hassle, I was
> able to set heavy error reporting on a single FastCGI pool and not
> across everything. So nice. So so nice!
>
> I've already migrated everything I do to use php-fpm. It's awesome. I
> cannot praise you enough, Andrei.
>
> Quick question: when I do php-fpm reload or php-fpm restart, does it
> physically kill the existing engines and someone could get a 502 bad
> gateway during that little bit of time, or does it somehow do it in a
> graceful fashion? It seems to restart so quick (and reload even
> quicker) that it doesn't even trip up my healthchecks so far... I'm
> wondering if it's doing it gracefully on purpose or it just works
> really fast.

"reload" does send SIGUSR2 to master process.

When SIGUSR2 is sent to master process it does gracefull reload and would
never cause requests to be aborted in the middle, unless some workers have
been unable to react on signals within process_control_timeout.

"restart" is composed from "stop" and "start", which may cause aborted
requests to happen some times.


--
Andrei Nigmatulin
GPG PUB KEY 6449830D

Now I lay me down to sleep(3)
Pray the OS my core to keep
If I die before I wake
Pray the Disk my core to take

mike503

unread,
Apr 9, 2008, 6:17:09 PM4/9/08
to highload-php-en
Minor annoyance:

cnisnlinux01:/usr/src# php-fpm stop
Shutting down php_fpm cat: /var/run/php-fpm.pid: No such file or
directory
kill: usage: kill [-s sigspec | -n signum | -sigspec] pid |
jobspec ... or kill -l [sigspec]
done
cnisnlinux01:/usr/src#

if the pid file doesn't exist, it gives some nasty error.

i guess it would be just some sort of simple -f check in the php-fpm
shell script? if no pid file, say "hey no pid file - is it even
running?" or something :)

Andrei Nigmatulin

unread,
Apr 10, 2008, 12:51:40 PM4/10/08
to highloa...@googlegroups.com

Yes, this to be added into next release.

Reply all
Reply to author
Forward
0 new messages